Access prediction service serving explainable risk scores
A process, system and medium for detecting anomalous authentication requests to a protected resource during an authentication journey, in order to regulate step-up authentication are described. The process includes obtaining request features from the authentication request that triggered the authentication journey. The process includes processing, by an ensemble of Machine Learning (ML) models and a set of rule-based heuristics, a set of features based on the request features, the set of features associated with a userID. The process includes deriving risk sub-scores for each ML model and each heuristic. The process includes deriving a risk score based on the risk sub-scores. The process includes determining that the risk score exceeds an explanation-triggering threshold. The process includes providing, to a node in the authentication journey, the risk score with an explanation of the risk score. The system and medium are configured to execute the process, and configured to provide the explanation.
1. A method for detecting anomalous authentication requests to a customer endpoint during an authentication journey, in order to conditionally invoke step-up authentication, the method comprising:
obtaining request features from an authentication request that triggered the authentication journey;
processing, by an ensemble of Machine Learning (ML) models and a set of rule-based heuristics, a set of features based on the request features, the set of features associated with a userID, the ensemble of ML models comprising unsupervised learning models including at least one k-mode clustering model and at least one encoder-decoder model, each cluster member of the at least one k-mode clustering model representing a distinct userID;
deriving a risk sub-score for each ML model from the ensemble of ML models and each heuristic from the set of rule-based heuristics, the risk sub-score for the at least one k-mode clustering model being based on a set of request features compared to a k-mode cluster-mode of the at least one k-mode clustering model;
generating an explanation of at least one risk sub-score;
deriving a risk score based on the risk sub-score for each ML model from the ensemble of ML models and each heuristic from the set of rule-based heuristics;
determining that the risk score exceeds an explanation-triggering threshold; and
providing, to a node in the authentication journey, the risk score with an explanation of the risk score,
whereby the risk score and the explanation can be used to determine whether to invoke step-up authentication.
2. The method of claim 1 , wherein the at least one encoder-decoder model includes at least a variational autoencoder (VAE), and the explanation of the at least one risk sub-score includes an explanation for the risk sub-score of the VAE.
3. The method of claim 1 , wherein the set of rule-based heuristics includes at least 4 of:
Impossible traveler,
Credential stuffing,
Brute force,
Suspicious IP, or
Automated User Agent (aka bot).
4. The method of claim 1 , wherein the set of features includes at least 3 or more request features of:
geographical country from which the authentication request originated,
geographical city from which the authentication request originated,
device type from which the authentication request originated,
day of week on which the authentication request was received,
part of day in which the authentication request was received,
OS type from which the authentication request was sent, or
user agent from which the authentication request was sent.
5. The method of claim 4 , wherein the set of features includes at least 5 or more of the request features.
6. A non-transitory computer-readable medium configured with instructions to detect anomalous authentication requests to a protected resource during an authentication journey in order to conditionally invoke step-up authentication, the instructions, when executed, implement a method comprising:
obtaining request features from an authentication request that triggered the authentication journey;
processing, by an ensemble of Machine Learning (ML) models and a set of rule-based heuristics, a set of features based on the request features, the set of features associated with a userID, the ensemble of ML models comprising unsupervised learning models including at least one k-mode clustering model and at least one encoder-decoder model, each cluster member of the at least one k-mode clustering model representing a distinct userID;
deriving a risk sub-score for each ML model from the ensemble of ML models and each heuristic from the set of rule-based heuristics, the risk sub-score for the at least one k-mode clustering model being based on a set of request features compared to a k-mode cluster-mode of the at least one k-mode clustering model;
generating an explanation of at least one risk sub-score;
deriving a risk score based on the risk sub-score for each ML model from the ensemble of ML models and each heuristic from the set of rule-based heuristics;
determining that the risk score exceeds an explanation-triggering threshold; and
providing, to a node in the authentication journey, the risk score with an explanation of the risk score,
whereby the risk score and the explanation can be used to determine whether to invoke step-up authentication.
7. The non-transitory computer-readable medium of claim 6 , wherein the at least one encoder-decoder model includes at least a variational autoencoder (VAE), and the explanation of the at least one risk sub-score includes an explanation for the risk sub-score of the VAE.
8. The non-transitory computer-readable medium of claim 6 , wherein the set of rule-based heuristics includes at least 4 of:
Impossible traveler,
Credential stuffing,
Brute force,
Suspicious IP, or
Automated User Agent (aka bot).
9. The non-transitory computer-readable medium of claim 6 , wherein the set of features includes at least 3 or more request features of:
geographical country from which the authentication request originated,
geographical city from which the authentication request originated,
device type from which the authentication request originated,
day of week on which the authentication request was received,
part of day in which the authentication request was received,
OS type from which the authentication request was sent, or
user agent from which the authentication request was sent.
10. The non-transitory computer-readable medium of claim 9 , wherein the set of features includes at least 5 or more of the request features.
11. A system, including a processor operatively connected to a non-transitory computer-readable medium with instructions that, when executed, perform steps of a method for detecting anomalous authentication requests to a protected resource during an authentication journey in order to conditionally invoke step-up authentication, the method comprising:
obtaining request features from an authentication request that triggered the authentication journey;
processing, by an ensemble of Machine Learning (ML) models and a set of rule-based heuristics, a set of features based on the request features, the set of features associated with a userID, the ensemble of ML models comprising unsupervised learning models including at least one k-mode clustering model and at least one encoder-decoder model, each cluster member of the at least one k-mode clustering model representing a distinct userID;
deriving a risk sub-score for each ML model from the ensemble of ML models and each heuristic from the set of rule-based heuristics, the risk sub-score for the at least one k-mode clustering model being based on a set of request features compared to a k-mode cluster-mode of the at least one k-mode clustering model;
generating an explanation of at least one risk sub-score;
deriving a risk score based on the risk sub-score for each ML model from the ensemble of ML models and each heuristic from the set of rule-based heuristics;
determining that the risk score exceeds an explanation-triggering threshold; and
providing, to a node in the authentication journey, the risk score with an explanation of the risk score,
whereby the risk score and the explanation can be used to determine whether to invoke step-up authentication.
12. The system of claim 11 , wherein the at least one encoder-decoder model includes at least a variational autoencoder (VAE), and the explanation of the at least one risk sub-score includes an explanation for the risk sub-score of the VAE.
13. The system of claim 11 , wherein the set of rule-based heuristics includes at least 4 of:
Impossible traveler,
Credential stuffing,
Brute force,
Suspicious IP, or
Automated User Agent (aka bot).
14. The system of claim 11 , wherein the set of features includes at least 3 or more request features of:
geographical country from which the authentication request originated,
geographical city from which the authentication request originated,
device type from which the authentication request originated,
day of week on which the authentication request was received,
part of day in which the authentication request was received,
OS type from which the authentication request was sent, or
user agent from which the authentication request was sent.